100
LNG
INDUSTRY
MARCH
2016
Industry tolerance
While the need for two infrastructures has admittedly
been a necessity in the past, it is probably most accurate
to say that users have ‘tolerated’ this situation rather than
‘preferred’ it. This is particularly true for IT departments
tasked with managing the computing, software, network,
and security infrastructures for a plurality of systems.
Furthermore, it is also true for the machinery engineers in a
typical plant, who must convince their IT departments that
they need a separate infrastructure that often consists of
different computers, operating systems, client applications,
firewalls, remote access environments, security models,
etc. Both stakeholders are inconvenienced in multiple
ways, and both would like a solution that is less expensive
to deploy and sustain by sharing, rather than duplicating,
infrastructure.
Convergence
There has long been an appetite to converge both process
and vibration data because the interaction between the
machine and the process surrounding it is inevitable.
Process conditions can adversely impact a machine, and
failures are often due to external influences rather than
normal wear and tear. For instance, in the case of pumps, it
might be process conditions that lead to cavitation; in the
case of compressors, it might be process conditions that
lead to a surge; and, in the case of a turbine generator, it
might be changing steam conditions that result in excessive
differential expansion and a mechanical rub. This means
that the ability to correlate process and vibration data is
frequently necessary when diagnosing problems in rotating
machinery – particularly critical machinery that is an
integral part of the process flow. Various cumbersome, and
usually expensive, methods of integrating machinery and
process data have existed for years, but often meant that
the data had to be replicated in both systems. This lead to
the inevitable ‘two versions of the truth’ that never agreed
precisely with one another. Until recently, the technology
had simply not advanced sufficiently to enable convergence
of these types of disparate data into a single repository. As
a result, separate repositories could share data with one
another, but were still just that – separate repositories.
Meanwhile, technology has
progressed at a staggering rate
along the familiar Moore’s Law
trajectory (Figure 2). Indeed, as
process data historians have
pushed the envelope of data
collection speeds, they now
easily surpass the
1 million tag/sec. mark and
continue to climb even higher,
with speed constraints primarily
imposed by the ability of hard
drives to write to their platters.
The advent of affordable
solid-state drives and raid
arrays has circumvented even
this constraint, and the speed
limit is climbing ever higher.
100 000 x 10 =
1000 x 1000
Process historians evolved
along a path where many
thousands of points needed to
be scanned, but at relatively
slow rates. Update rates of
1 sec. for process data are often
considered to be extremely
Figure 1.
Most data in the plant flows from purpose-built
controllers/monitors into the Distributed Control System (DCS),
and then into the process historian. However, the special needs of
vibration analysis software have historically required a stand-alone
‘silo’ that uses its own infrastructure, separate from the process
historian.
Figure 2.
Microprocessor transistor counts 1971 – 2011 and Moore’s Law.
